Research On Automatic Recognition And Time-Stamp Of Power Quality Disturbances

Electric power plays a key role in industrial production and daily life. In recent years, the load structure has varied greatly. A great deal of nonlinear loads, fluctuant loads and impactive loads have swarmed into power system. They have improved the productive efficiency and the living quality, but they have also polluted the power system and decreased the quality of power supply. On the other hand, with the development of computer science and semiconductor technology, automatic and intelligent equipments controlled by them are widely used in industry. These equipments are very sensitive to power quality disturbances. The loss caused by power quality is enormous. Therefore, power quality has received more and more attention in recent years. Research on the detection, analysis and control of power quality becomes a hotspot problem in power system realm.The ultimate purpose of study on power quality is to decrease its effects on sensitive loads. Monitoring of power quality provides reliable evidences to solve the problem and it is the basis for power quality control. It is very essential to set up monitoring system to detect, evaluate and classify power quality. Power quality indices calculating methods and the analysis methods are the bassis for power quality monitoring. These methods should be checked by their accuracy, correctness and feasibility. There are two kinds of power quality:steady state power quality and transient power quality. There are many differences on detection, analysis and evaluation methods between them. The development of digital signal processing technology provides a great deal of methods to detect and analyze power quality problems. Fourier transform is often used for steady state power quality and it has FFT for fast calculation. But the methods used to detect and analyze transient power quality are complicated. Intantaneous voltage disturbances such as voltage sag, swell and interruption are transient power quality problems which has big effects to consumers. Reach on intantaneous voltage disturbances is significancy.Detection and time-stamp for instantaneous power quality provide proofs to estimate the reasons of power quality disturbances. It becomes a hotspot in power quality research. Instantaneous power quality signal often has singularity at the start and the end. This characteristic is used to detect instantaneous power quality and can be used to make time-stamp. Detection and time-stamp methods often used are time-domain disturbance detection method, neural network, wavelet transform, Teager energy operator, mathematical morphology transform and Hilbert-Huang transform, etc. Singular point is hard to be detected if the original signal has noise. It is a key point to detect the singular point form a noisy signal. And it can be used to evaluate whether a detection method is good or not.Correct classification of power quality can provide proofs to solve the problem. Therefore, it is important to study the methods of classifying power quality. The methods contain two aspects:the extraction of feature vector and the classifier. The feature vector should represent the original signal uniquely and its size should be as small as possible. The methods often used to extract feature vector are wavelet transform, S-transform, Hilbert-Huang transform and the ones derived from them. Classifier is designed based on artificial intelligence. Artificial neural network, support vector machine, fuzzy logic, expert system and fuzzy-expert system are often used to design classifier.In this thesis, instantaneous power quality is the main object. Research is devoted on the detection and time-stamp of instantaneous power quality disturbances, the classification of power quality disturbances and the design of an integral power quality monitoring system. The main contributions of the dissertation are as following:(1) An automatic classification method of power quality disturbances is proposed based on wavelet energy distribution and BP neural network. Wavelet is used to perform multi-resolution to the original signals. The energy to every level is calculated using wavelet coefficients and the energy distribution is got by combining the energies into a vector. Also, the energy distribution of a standard signal is calculated. The difference of the two energy distribution is used as the feature vector. A three-layer BP neural network is used as the classifier. The output of the BP neural network gives out the type of the disturbance. Signals used to train and simulate neural network should be representative. That is the start point and the duration of the power quality disturbances should be stochastic. Also, the signals should contain noise component to examine the validity of the proposed method. Simulation results indicate that the identification rate of the power quality disturbances is high under noisy conditions. The method proposed in this thesis is proved to be valid.(2) A method to detect the transition points of instantaneous voltage disturbances based on auto-regressive model (AR model) is proposed. Every sampling data of the original signal is estimated by AR model. The residual sequence is got for every sampling point using the original value and the estimation. Transition points are then detected by allocating the time instants where residuals are prominent and they are the instant where the instantaneous voltage disturbance occur and end. So time-stamp can be made according to transition points. In the process of disposal, the data sequence is first subdivided into overlapping and fixed-size segments. Parameter estimation is applied to every data segment using AR model. The AR model keeps constantly during estimating the values in the segment, but it is time-varying between two data segments. The rank is stable during the whole estimation. Simulation results indicate that the proposed method can locate the transition points of instantaneous disturbances correctly and can get correctly result even when the original signal is polluted by noise and harmonic. Power quality caused by transformer energizing is studied through experiment. Transformer energizing can cause inrush current. The inrush current contains 2nd harmonic, which has effects to power system. Transformer energizing can also cause voltage sag in the bus and the continuance time is longer because of the voltage increasing gradually. Voltage sag caused by transformer energizing is solved by the proposed method. For the start point, the time-stamp is very exact.(3) A detection and time-stamp method of instantaneous voltage disturbances based on Hankel matrix singular value decomposition is proposed. Firstly, Hankel matrix is constructed using the time series of voltage signal. Then the singular value decomposition is executed to the Hankel matrix. The decomposition signals are calculated by the decomposition results. The decomposition signals are linear superposition. In some decomposition signals, the transition points show prominence and the instant where the intantaneous voltage disturbance start and end can be got. So time-stamp can be made according to transition points. At each side of the start point in the original signal, one cycle signal is taken out to calculate the fundamental component by FFT. The type of the voltage disturbance can be got by the relationship between the fundamental components and the indices can also be calculated. Signals contain single disturbance, hybrid signals and real signals are used in simulation. The results show the correctness and efficiency of the proposed method. The applicability and concentricity of locating transition point is compared between wavelet and the proposed method, the result indicates that the proposed method is superior to wavelet. The computation rate is moderate and the proposed method is fit for online analysis.(4) Integral power quality monitoring system is designed in this dissertation. The system contains three subsystems:terminal system, data server system and communication system. Functional analysis is made for every subsystem. Functional structures and flowcharts are given. Software is designed by LabVIEW. The terminal system contains two parts:hardware system and software system. The hardware system contains transformer, signal conditioning circuit, data acquisition card and industrial computer, etc. The software system contains data acquisition module, power quality indices calculation module, power quality data storage module and power quality statistics module, etc. Hardware of the data server system can use a general computer. The data server system's software is the main part. Its function is to perform the statistics of power quality data. It can display data friendly or give reports with the desire of users. The terminal system is usually set up in central substation where communication channels can be used. So the hardware of communication system is not introduced. Communication software running at the terminal system and data server system is design based on DataSocket. Information feedback is used between both sides to ensure the reliability of communication. Instantaneous voltage disturbances experiment system is set up. The method to detect instantaneous voltage disturbances based on Hankel matrix singular value decomposition is used in the program. Testing results indicate that the method is correct.